PAX5 and B-cell neoplasms: transformation through presentation

What ever Pax5 soweth…

The salient features of B-cell development are immunoglobulin gene rearrangements, class switching recombinations (CSRs) and somatic hypermutations (SHMs). Collectively, these processes account for the diversity of antibodies, as well as for affinity maturation of antibodies in germinal centers. However, the propensity of B cells to undergo genetic alterations has unintended consequences – that is, a high frequency of oncogenic events. These events clearly inform lymphomagenesis: translocated and otherwise altered genes inevitably turn out to be oncogenes playing an important role in tumor initiation. For example, Burkitt’s and some diffuse large B-cell lymphomas (DLBCLs) carry the t(8;14) translocation that places the MYC proto-oncogene under control of the Igh enhancer (for a recent review see [1]).

Another transcription factor affected by translocations is Pax5. Pax5 belongs to a family of nine nuclear proteins that are thought to control the tissue-specific transcription necessary for many types of cell differentiation. Pax5 is the only family member found within the hematopoietic system, and its expression is restricted to certain stages during B-cell differentiation [2]. It is largely absent from multipotent progenitors and common lymphoid progenitors [3], but its expression is initiated in pre-pro-B cells and then maintained throughout subsequent stages of B-cell development [4] before it is downregulated in plasma cells [5]. However, Pax5 is unique among other lineage-specific transcription factors in that it is both a driving force behind and one of the primary beneficiaries (or perhaps sufferers) of genomic alterations.

Indeed, one of its important targets is the Rag-2 gene encoding recombination-activating protein 2 [6,7]. During V_H–DJ_H recombination, the RAG1/RAG2 complex introduces double-stranded DNA breaks that ultimately rejoin to generate functional heavy chain cassettes. The same process is thought to underlie illegitimate recombination, during which Ig heavy chain enhancers will be juxtaposed to one of the B-cell oncogenes such as Myc, Bcl2, Bcl1 and so on. In addition to activating the Rag2 gene, Pax5 occupies V_H genes in early human and mouse B-lineage cells, induces Igh locus contraction [8], and physically interacts with the RAG1/RAG2 complex, thus promoting recombination events [9]. Importantly, the V_H–DJ_H recombination cannot happen in the presence of lysine-9-methylated histone H3, and the removal of this inhibitory signal in B cells is an important function of Pax5 [10]. Consequently, only a minor fraction of lymphocytes in Pax5-null mice have completed V_H–DJ_H rearrangement and most are arrested at the earlier, pro-B stage [11]. Recent data indicate that Pax5 might also play a role in light chain locus contraction [12]. Additionally, Pax5 is known to be required for the expression of Igα [13], a sIgM-associated signaling molecule, which is the key component of the B-cell receptor (BCR; reviewed in [14]).

CSR, which might also contribute to oncogenic translocations, and SHM are less well-understood processes. However, recent work has implicated activation-induced cytidine deaminase (AID) in both phenomena. Interestingly, Pax5 is known to activate AID [15,16] and can therefore contribute to CSR and SHM also. Overall, while Pax5 might be a guardian of B-cell identity [2], it is a curiously permissive one with respect to genome integrity.

…he shall also reap

The cruel irony is that while Pax5 loyally acts on behalf of B cells to ensure their genetic diversity, it is one of the first to suffer from that diversity. Indeed, either an erroneous V_H–DJ_H rearrangement or CSR are likely responsible for the relatively rare [17], but persistent, t(9;14) (p13; q32) translocation [18–20] associated with aggressive B-cell non-Hodgkin lymphomas [21]. This translocation usually brings together the coding sequences of PAX5 and a strong enhancer from the Igh locus, ensuring robust transcription of the translocated PAX5 allele. In addition to disturbing Pax5-guided gene expression during lymphocyte differentiation, the translocation can potentially lead to failed PAX5 repression at the onset of plasma cell differentiation [2].

In addition to genomic rearrangements, the Pax5 gene is also affected by SHM. In the high-profile 2001 study, more than 50% of human germinal center-derived DLBCLs were found to contain mutations in the proto-oncogenes PIM, MYC, RhoH/TTF (ARHH) – and Pax5 [22]. In the latter, SHM were identified downstream of both transcription initiation sites, predominantly around exon 1B, at the frequency of 0.15 per 100 bp. Although three missense mutations were detected within the short coding portion of exon 1B, most of the changes involved 5´ untranslated and other noncoding sequences [22]. This distribution implied that the main consequence of these gain-of-function SHMs is transcriptional activation of Pax5 and, plausibly, B lymphomagenesis.

Two studies using shRNA-mediated gene knockdown appear to corroborate this notion. The RNAi experiments were performed in vitro on two cell lines, SUDHL-4 and -6, belonging to the germinal center B cell-like subclass of DLBCL [23]. In the first study, the authors achieved knockdown of the key downstream effector of Pax5 Igα (also known as CD79a) [24]; in the second study, from our laboratory, the Pax5 gene itself was targeted [25]. Under both experimental conditions, decreases in cell accumulation were observed, supporting the oncogenic role of Pax5. However, other recent data suggest that the picture might be much more complicated.

Acute lymphoblastic leukemia versus non-Hodgkins lymphoma: why does it matter?

The surprising findings came from the analysis of B-progenitor acute lymphoblastic leukemia (B-ALL), a common pediatric malignancy. It turned out that Pax5 is important for B-ALL pathogenesis – but not in the way one would expect. In their December 2006 Blood paper, Bousquet et al. report a novel t(7;9)(q11;p13) translocation in two B-ALL patients, resulting in the fusion of the Pax5 and elastin (ELN) genes [26]. Fusions between Pax5 and various translocation partners are not new. For example, the previously described t(9;12)(q11;p13) translocation generates a fusion transcription factor containing DNA-binding domains of both Pax5 and ets variant gene 6 (ETV6)/translocation, ets, leukemia (TEL) [27]. What is interesting regarding the Pax5–ELN fusion is that not only is it loss-of-function, but it also acts as a dominant-negative isoform in transient expression assays [26].

In parallel, Mullighan et al. embarked on the genome-wide analysis of B-ALL using high-resolution SNP arrays and direct genome sequencing. Surprisingly, the two most commonly mutated genes were Pax5 and its upstream regulator, early B cell factor 1 (EBF1) [28]. Again, the mutations identified were, without exception, loss-of-function: monoallelic deletions of EBF1 and Pax5, cryptic transloctions removing the Pax5 activation domain, frameshift mutations in Pax5, and so on. Importantly, both transient expression assays and the analysis of PAX5 target genes confirmed that Pax5 function is consistently reduced (although not completely abolished) in B-ALL cases. The authors conclude that subtle changes in the dosage of PAX5, rather than its overexpression, underlie the pathogenesis of ALL [28]. While this might certainly be the case, an important question is whether gain-of-function mutations in the Pax5 gene play any role in B lymphomagenesis at all.

One possible answer is that it does, but only at certain stages of B-cell development. Most DLBCL and other non-Hodgkin lymphomas are derived from mature B-cells, which express BCR and have often undergone maturation in germinal centers [29]. In contrast, most of B-ALL (so-called A1 and A2 types) are derived from immature pro- or pre-B-cells lacking BCR [30]. Given the contribution of Pax5 to BCR signaling (see above), it seems plausible that Pax5 can only exert its oncogenic effects when this pathway is functional. However, an experimental verification of this simple model has been slow in coming.

A man or a mouse?

Needless to say, dissecting the causative role of putative oncogenes in spontaneous human tumors can be tricky. However, the knock-out/knock-in mouse technology tends to be helpful in this regard. With this goal in mind, the entire Pax5-coding sequence has recently been knocked into the Igh locus, thus recreating the t(9;14) translocation in the mouse. The animals did develop overt neoplasms, but rather unexpectedly developed T, and not B lymphomas [31]. Apparently, the germline insertion of the Pax5 gene and its ensuing activation throughout the lymphoid system arrests T-cell development and leads to malignant T lymphomas. This rather artificial system is nevertheless reminiscent of idiopathic activation of Pax5 in forebrain-derived astrocytomas [32] and a subset of highly malignant neuroblastoma cell lines [33].

However, to assess its role in B lymphomagenesis, new approaches were in order. Our laboratory took advantage of a previously described cell line (Myc5) derived from a p53-null, Myc-overexpressing lymphoma [34]. The original Myc5 tumor was Pax5-positive, but upon culturing in vitro, Pax5 expression was spontaneously extinguished and B-cell markers were lost. Instead, the cultured cells readily engulfed latex beads and provided T-cell help such as bona fide macrophages [35]. Nevertheless, upon re-expression of Pax5, Myc5 neoplastic cells expressed both surface IgM and surface IgD [36], which is characteristic of mature B cells (the so-called fraction F [37]). We reasoned that this system is well-suited to test the role of BCR signaling in Pax5-assisted lymphomagenesis.

Antigen presentation & neoplastic transformation

To analyze the immediate effects of Pax5 on B lymphomagenesis, we generated Myc5 subclones expressing the hydroxytamoxifen (4OHT)-inducible variant of Pax5. Upon re-injection into syngeneic recipients, these clones formed much larger tumors in 4OHT-treated than in control mice [25]. To reveal the mechanism of Pax5-dependent lymphomagenesis, we profiled mRNAs in Pax5-sufficient and -deficient tumors. We discovered that the repertoire of Pax5-activated genes was highly enriched in components of BCR signaling: Igα, CD19 antigen, B-cell linker (BLNK), Bruton agammaglobulinemia tyrosine kinase (Btk), and so on. In addition, two crucial inhibitors of BCR signaling, CD22 and paired-Ig-like receptor B, were repressed. What role do all these genes play in B lymphomagenesis?

To answer this question, we used the fusion protein containing the cytoplasmic regions of Igα and Igβ and constitutively targeted this to the plasma membrane [38]. Transduction of parental Myc5 cells with the Igα/β cassette resulted in strong stimulation of tumor growth. Conversely, forced expression of CD22, which antagonizes BCR activity [39], completely canceled the effects of Pax5 on tumor growth. Similar attenuation of tumor growth could be achieved through the use of pharmacological inhibitors of Syk, a BCR-associated tyrosine kinase [25].

Future directions: Pax5 & BCR as therapeutic targets

It has been proposed that BCR signaling (ligand-dependent or -independent) can promote survival of neoplastic B-lymphoma cells, for instance, via activation of antiapoptotic Mcl1 [40]. Increased BCR signaling was indeed observed in cultured B-lymphoid cells [41], and appeared to contribute to their growth [24]. However, the role of BCR signaling in tumor growth in vivo only has been inferred from circumstantial evidence: its persistent expression in most non-Hodgkin lymphomas, the retention of functional Ig alleles in lymphomas with IgH translocations and the discovery of autoreactive BCR in some neoplasms (reviewed in [42]).

To determine whether our experimental findings might be applicable to spontaneous human lymphomas, Dr Teresa Marafioti’s group stained tissue arrays of human DLCBLs for both PAX5 and phosphorylated BLNK, the key adaptor protein of activated BCR. Her group observed that all DCBCL tumors stained positively for PAX5 and approximately half were also positive for phosphorylated BLNK. Thus, BCR is likely to be constitutively activated in at least some human B-cell tumors, attesting to the important causative role of PAX5 in B lymphomagenesis.